OpenAI’s ChatGPT uses site: operators to improve search quality

Image credit: Search Engine Journal

OpenAI’s ChatGPT is increasingly using specific ‘site:’ search operators to curate information from higher-quality sources, enhancing response accuracy and reducing spam globally.

This strategic shift in how ChatGPT retrieves and processes information highlights OpenAI‘s commitment to refining its models and improving the overall user experience by combating low-quality data.

Analysis of these fan-out queries offers researchers insight into the AI’s search behavior, revealing the types of information it prioritizes and its expected sources, according to experts.

Lily Ray, a senior director at Amsive Digital, said the number of retrieved URLs in ChatGPT’s responses is growing, while the number of unique domains cited per response is simultaneously decreasing, suggesting a more focused retrieval strategy.

Around August, the ‘Luna’ variant of ChatGPT 5.6, a more cost-effective model, became the default for Free and Go users, impacting the information experience for the majority of the platform’s user base, Search Engine Journal reported.

Olivier de Segonzac, an SEO consultant, said monitoring these real-time fan-out queries provides a unique window into ChatGPT’s evolving information-seeking patterns.

Tools such as Peec AI, Profound, the Resoneo plugin and FanoutFox allow researchers to observe ChatGPT’s search behavior and the sources it consults in real time.

OpenAI has been actively working to refine its information retrieval process, which analysts at Search Engine Land described as a direct method to improve the quality of retrieved sources and combat the proliferation of low-quality information.

Source: Search Engine Journal